Browsing by Subject "State Estimation"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
Item Multiple Target Tracking Using Random Finite Sets(2021-01) Siew, Peng MunMultiple target tracking (MTT) plays a crucial role in guidance, navigation, and control of autonomous systems. However, it presents challenges in terms of computational complexity, measurement-to-track association ambiguity, clutter, and miss detection. The first half of the dissertation looks into multiple extended target tracking on a moving platform using cameras and a Light Detection and Ranging (LiDAR) scanner. A Bayesian framework is first designed for simultaneous localization and mapping and detection of dynamic objects. Two random finite sets filters are developed to track the extracted dynamic objects. First, the Occupancy Grid (OG) Gaussian Mixture (GM) Probability Hypothesis Density (PHD) filter jointly tracks the target kinematic states and a modified occupancy grid map representation of the target shape. The OG-GM-PHD filter successfully reconstructed the shape of the targets and resulted in a lower Optimal Sub-Pattern Assignment (OSPA) error metric than the traditional GM-PHD filter. The second MTT filter (Classifying Multiple Model (CMM) Labeled Multi Bernoulli (LMB)) is developed to leverage class-dependent motion characteristics. It fuses classification data from images to point cloud and incorporates object class probabilities into the tracked target states. This allows for better measurement-to-track associations and usage of class-dependent motion and birth models. The CMM-LMB filter is evaluated on KITTI dataset and simulated data from CARLA simulator. The CMM-LMB filter leads to a lower OSPA error metric than the Multiple Model LMB and LMB filters in both cases. The second half looks into sensor management for MTT using a sensor with a narrow field of view and a finite action slew rate. The sensor management for space situational awareness (SSA) is chosen as an application scenario. Classical sensor management algorithm for SSAtends to only consider the immediate reward. In this dissertation, deep reinforcement learning (DRL) agents are developed to overcome the combinatorial increase in problem size for long-term sensor tasking problems. A custom environment for SSA sensor tasking was developed in order to train and evaluate the DRL agents. The DRL agents are trained using Proximal Policy Optimization with Population Based Training and are able to outperform traditional myopic policies.Item Position Estimation Using Magnetic Fields(2018-11) Madson, RyanThis thesis develops position estimation systems based on magnetic fields and addresses a number of challenges related to making such systems accurate and robust for real-world applications. The thesis first addresses one-dimensional position estimation using the measurement of piston position inside a cylinder as a benchmark application. The piston is equipped with a permanent magnet and one or more magnetic sensors are embedded on a compact circuit board located on top of the cylinder. Due to large distances between the moving piston and the stationary sensor, the magnetic field as a function of piston position is highly nonlinear. This magnetic field is modeled either analytically or emperically and a nonlinear estimation algorithm, namely the truncated interval unscented Kalman Filter (TIUKF), is utilized for real-time estimation of the position of the piston. Piston position estimation can be useful on hydraulic actuators, pneumatic actuators, IC engines, and a number of other cylinder piston products. The developed estimation algorithm is implemented experimentally on a microprocessor. A compact sensor board containing sensors, the microprocessor, and other components is developed. The developed position estimation system is first evaluated experimentally on pneumatic actuators. The estimation system performs well and an estimation accuracy better that 1% is achieved on pneumatic actuators with stroke lengths of 5 cm and 10 cm. Next an auto-calibration system is developed in order to enable the sensor board to estimate position accurately when installed on new cylinders. Small misalignments and offsets in location can occur on each installation. The new auto-calibration method allows the position estimation system to perform robustly and accurately by identification of new parameters on each installation. This auto-calibration is done without requiring any additional external reference position sensors. A significant challenge to magnetic field based position estimation comes from disturbances due to unexpected ferromagnetic objects coming close to the sensors. A new disturbance estimation method based on modeling the magnetic disturbance as a dipole with unknown location, magnitude, and orientation is developed. A TIUKF is used to estimate all the parameters of this unknown dipole, in addition to estimating piston position from nonlinear magnetic field models. Experimental data from a pneumatic actuator is used to verify the performance of the developed estimator. Experimental results show that the developed estimator is significantly superior to a linear magnetic field model based disturbance estimator. It can reliably estimate piston position and the unknown dipole parameters in the presence of a variety of unknown disturbances. Next the estimation system is implemented for a large hydraulics actuator used on construction machines. The ferromagnetic material of the hydraulic cylinder leads to significant hysteresis, since this material is magnetized and demagnetized repeatedly with the motion of the magnet. A method to model and compensate for the hysteresis in the system is developed. In particular, a modified Preisach model and associated estimation algorithm developed is shown to provide excellent performance. An accuracy better than 2\% is achieved on the large-stroke hydraulic cylinder in spite of significant hysteresis. Finally, the one-dimensional position estimation tools are extended in an attempt to enable 3D position estimation of a magnet. The objective is to estimate magnet position in real-time from a moving sensor board in the neighborhood of the magnet. Applications for this 3D position estimation system include a breast cancer surgery application in which a small magnet can be used to mark tumor location. The significant challenges in the 3D position estimation application are handled by using an accelerometer and gyroscope in addition to magnetic sensors for orientation estimation, by using a particle filter for the estimation task, and by using a neural network for modeling the functional relationships between magnetic field and 3D position and orientation. While the developed system provides reasonable experimental performance, further work with more sensitive magnetic sensors and a better reference 3D position sensor for modeling are needed.Item Real Time Angle Of Attack Estimation For The Hycube Flight Vehicle In Gps-Denied Re-Entry(2024-01) Vedvik, SophiaWith global interests in the study of hypersonic flow, large research efforts have gone towards collecting statistically significant amounts of high speed flow data at low costs. CubeSats are proving to be an economical testing platform for a variety of scientific experiments, where valuable hypersonic data can be collected and relayed upon re-entry to Earth. However, due to the budget, volume, and power constraints of CubeSats, many of the on-board sensors, including inertial measurement units (IMUS), have decreased accuracy. For purposeful data collection to occur, the sensors on-board typically work in conjunction with robust synthetic air data algorithms. To back out useful data on the vehicle's response during re-entry, the angle of attack of the vehicle must be estimated with one of such algorithms. This work proposes using an Extended Kalman Filter (EKF) which fuses an attitude determination algorithm with low-grade IMU angular rates and measurements of Earth's magnetic field. But in the case of re-entry, the vehicle will become deprived of Global Positioning System (GPS) data, which is required to obtain estimates of the Earth's magnetic field that work in conjunction with magnetometer magnetic field measurements. Thus, after developing the EKF framework, this work will perform a trade study to analyze ways in which Earth's magnetic field can still be a viable method to aid low-grade IMU attitude estimates. The trade study environment is modeled after the Hypersonic Configurable Unit Ballistic Experiment (HyCUBE), a project in development at the University of Minnesota that is leveraging the CubeSat form factors to collect valuable hypersonic flow data upon re-entry. Future work and improvements to the EKF, as well as the impact of this work will then be discussed.